Pre-Regulator For Low Voltage Supplies

There will be many times where it is desirable to use the P05 supply module from a higher voltage source. For example, if you want to add balanced inputs to a power amplifier, then you need a ±15V supply, but the amp's supply voltage will be much too high for the regulator ICs.

This project is about as simple as they come, and is very cheap to build. It is designed for exactly this purpose - to reduce the amplifier supply voltage to a safe value for regulator ICs. It's worth noting that even though the LM317/337 (for example) have a rated voltage differential of 40V (the actual voltage across the IC), this is really the absolute maximum voltage that should ever be applied. Ideally, it will be kept much lower, and I recommend a maximum input voltage of around 25V. This reduces device power dissipation dramatically, and ensures that the ICs are well within ratings at all times.

While the datasheets claim that these regulators can be used at higher voltage as long as the voltage differential between input and output is within the ratings, they will fail if the output is shorted, or even if the output caps are too large (acting as a 'temporary' short). Use of a pre-regulator minimises the chance of failure.

In some cases you can use the pre-regulator by itself (with reduced output voltage of course) as the power supply to opamp based circuits. You don't need regulation with the majority of opamp circuits and their power supply rejection is usually very good. If you can reduce the main supply ripple by 50-60dB, no noise will be audible. Naturally, this is something that you must test for yourself - there are too many possibilities in circuitry, and no one solution is suitable for all applications.

All the circuits shown can be improved, potentially dramatically, by adding current sources to replace the resistors. However, this increases the complexity, and makes a simple circuit less simple and far more difficult to put together. The idea is to provide a means to reduce high power amp rail voltages to something cleaner (less ripple) and more regulator friendly. The circuits are not intended to be regulators in their own right, but in some cases they may be all that you need.

The basic circuit is shown in Figure 1 and it is very simple indeed. You will need to make a few simple calculations to determine the resistor values, and these are explained below. There is also an 'enhanced' version that provides even better ripple and noise rejection - see below. Finally, there's a version using MOSFETs which some may prefer.

The circuit shown uses the 24V zener diodes (D1 and D2) to regulate the output voltage to a little under 24V. This is a perfectly safe input voltage for standard 3-terminal regulators, and using this circuit will provide even better regulation and supply noise rejection then normal. Using MJE3055 and 2955 transistors will allow for supply voltages up to 56V quite safely, but they will need to be mounted on a heatsink (with insulating washers). If you have a supply voltage of more than 56V, use transistors with a higher voltage rating.

In some cases it might be necessary to include a 10Ω 1W resistor in the GND connection from the main supply. This is something that you may need to experiment with if a ground loop (and subsequent buzz) causes a problem. In most cases you won't need it, but sometimes there will be no alternative. It can be useful to bypass the resistor with a 100nF polyester capacitor so that RF noise is properly grounded.

I suggest that R2 and R4 should be rated for 1/2W, and 1W zeners are recommended. Optionally, you can add a 100nF high frequency bypass cap in parallel with R2 and R4. Don't expect a difference though, unless RF interference is common in your area.

The only calculation is to determine the value for R1 and R3. First, measure the power amp supply voltage (V1). The resistor value is calculated to provide a maximum zener current of 20mA, and this will ensure sufficient base current for the pass transistors for up to 100mA or so output current at ±15V. If the current drain of your preamp is greater than 100mA, you'll need to allow for more base current for the series-pass transistors. Be careful that you don't reduce the resistance value to the point where the zeners dissipate more than 50% of their rated power or they will run far too hot.. For the following, I'll assume that the transistor's base current is 10mA, and we want 10mA zener current - a total of 20mA.

The dissipation in Q1 and Q2 may also be calculated, but you need to know the current drawn by the external circuits. For example, if the external circuitry draws 50mA, the transistor power dissipation is ...

That's it for the basic version - it could hardly be simpler. Your regulator ICs are safe, and have around 30 to 40dB less input ripple to contend with. This means that if the main supply rails have up to 6V of ripple at full load (as well as voltage variation due to varying current drain), this will be reduced to about 150mV total variation. The transistor base current will generally be less than 2mA for 50mA output (allowing for a transistor h_FE of 25 - 50).

If you need more output current you can use Darlington transistors, such as the TIP140 (NPN) and TIP145 (PNP) or their higher voltage versions if needs be. These can easily supply over 1A while allowing you to use the resistor values as calculated above. With a typical gain of around 1,000 you can even increase the resistor values if desired. Remember that the zener current should be set for about 20% of its rated maximum current (allow 10mA or up to a suggested maximum of 20mA for 24V zeners). To reduce zener dissipation, you can use 2 × 12V zeners in series. The pinouts for the TIP140/145 are the same as shown in Figure 4.

While the basic circuit shown already has quite good noise rejection, some applications might need the maximum possible noise rejection. If this is the case, you can use the version shown below. The Value of R1 and R3 are calculated exactly as before, but R1A and R1B are half the value calculated, and the same for R3A/ R3B. Power dissipation in each resistor is half that calculated above for the 'basic' version.

In this case, you'd use either a 50V or 63V cap, depending on which is easier to get and cheaper. The resistors (R1A/B and R3A/B) would be either 1.5k or 1.8k ohms. One is a bit lower than the total calculated value and the other is a bit higher, but either will be fine. Increasing the capacitor value will give even better noise rejection, but the supply will take a lot longer to reach full voltage. With the 100µF cap and 1.5kΩ resistors for R1A/B and R3A/B, it will take around 330ms before the output voltage stabilises.

Figure 2 - Enhanced Pre-Regulator Schematic

The added capacitor ensures that there is reduced ripple in the zener current, so the output voltage will also have lower ripple. You can expect an additional ripple voltage reduction of around 15dB with the values determined here (for a total of 45dB). Increasing the capacitor value will improve things further, but any regulator IC can easily handle the output of the circuit shown. As an added benefit, the output voltage is also relatively free of high-order harmonics, because the added capacitor acts as a low-pass filter.

C3 and C4 are optional. They do help, reducing high order harmonics further and reducing the overall ripple by about another 6dB. Whether you consider the added cost to be worthwhile is up to you - personally, I wouldn't bother because the caps will typically be mounted close to the zeners so will get hotter than normal and may have a reduced life. However, the AC ripple current is tiny so a bit of extra heat is probably not a major problem. As you would expect, the extra capacitance does increase the time before you have full output voltage.

As before, you can use Darlington transistors if more output current is needed, or if you want to use higher value resistors for better noise filtering with the same capacitance.

If you need a circuit that can either pass more current or that would benefit from the lowest possible ripple, a MOSFET is a good choice. The output voltage is less predictable because of the gate-source threshold voltage (it can vary by several volts, and depends on output current). This means that although you can get very high ripple rejection even with low value capacitor(s), the output voltage will vary with current. The circuit diagram is shown next. The suggested MOSFETs are rated for 100V, and more current than you will ever need. They can be replaced by almost anything else that you may have in your parts bin, provided that they have suitable ratings for your application.

Figure 3 - MOSFET Pre-Regulator Schematic

You still need to provide a reasonable current through the zener diode, and the calculations shown above are still required. The biggest difference is that the output voltage will be up to 5V lower than the zener voltage (D1, D2), so higher voltage zeners are used. There is no gate current, so no allowance is needed for the base current for a bipolar output. The second pair of zener diodes (D3, D4) is necessary to protect the MOSFET gates, which will be destroyed if the gate-source voltage exceeds 20V or so for any reason.

The current through the 24V zeners should not be less than 4-5mA, so R1A/B and R3A/B would be 2.7k for a ±56V input. The output voltage will be around 4-5V less than the zener voltage. Ripple can be expected to be less than 1mV, and does not change appreciably with output current. Ripple can be reduced further by increasing the value of C3 and C4 in parallel with the zener, but there is probably no point because output ripple will typically be less than that from the Figure 2 circuit.

Heatsink requirements are determined in the same way as for the other circuits shown. MOSFET dissipation is determined by the voltage across the MOSFET and the current though it, and for most preamps (under 100mA), and will usually be less than 5W.

Figure 3A - Simplified MOSFET Pre-Regulator Schematic

The MOSFET design can be simplified, as shown above. The zener diode current (D1, D2) is very low, but I've tested it with ±35V supplies, and the limits of the zener diodes won't be reached with any voltage under ±80V, a limit set by the MOSFETs. There are no calculations, and ripple rejection should be at least 50dB with the very simple arrangement shown. The output voltage is about ±23-23V with typical MOSFETs. The only thing you have to worry about is the MOSFET dissipation. With ±35V input and a 100mA load, the dissipation is around 1.3W, and that's not a lot of heat to get rid of. If the input voltage is 80V, the four resistors should be 0.5W (they will dissipate around 350mW).

This version has been installed in a recent project of mine, and it works perfectly. No calculations at all. You can even eliminate R1, C1 and R3, C2 (replace R2 and R4 with 10k). For voltages above ±60V, use 1W resistors. The ripple rejection is not as good, but it's still more than acceptable for the inputs to the regulators, being a lot less than you'd normally get with a separate rectified and filtered low voltage supply. For anyone who likes to 'keep it simple', you don't get much simpler than this!

Construction is non critical, and the resistors, zener and power transistors can be mounted on a tiny piece of Veroboard or similar. There are no stability issues, and you only need to make sure that the transistors have an adequate heatsink. Mounting to the chassis will normally be quite sufficient - even a steel chassis will keep the temperature well within limits. Remember that the transistor cases must be electrically isolated from the chassis, and silicone pads will be fine due to the low dissipation. Do not use tantalum capacitors in any of the circuits shown - they are the most unreliable caps ever made, and I don't recommend them for anything.

A suggestion for assembly is shown in Figure 4. This construction method will be quite acceptable for most applications. The transistor cases must be isolated from the heatsink with silicone washers. The assembly shown is for the Figure 1 circuit. It includes the optional resistor and capacitor that can be added if earth/ ground loop hum is experienced. The network may not work in all cases, and may require experimentation to obtain minimum hum levels.

The above does not include the additional capacitors shown in Figure 2 and can't be used for the Figure 3 circuit. If you want to use either of the alternative circuits, you'll need a piece of Veroboard or other prototyping board to mount the extra parts. A heatsink will be needed in almost all cases, unless your preamp only requires a very low current. With a 56V input and 20mA output, dissipation will be over 0.6W, so only the most basic heatsink is necessary - a piece of 1mm flat aluminium will suffice, or use the chassis. You need to determine the dissipation with your circuit (based on the current drain and main supply voltages) to work out how much heatsink you will need.

Connect to a suitable power supply - remember that the supply earth (ground) must be connected! When powering up for the first time, use 100 ohm to 560 ohm 'safety' resistors in series with each supply to limit the current if you have made a mistake in the wiring.

There is very little that can go wrong (other than wiring mistakes), so any fault you may find is easily rectified. Note that extreme care is needed against shorting the outputs of any of the circuits shown, as there is no short circuit protection and failure is almost guaranteed. The regulators have their own protections circuits, so a shorted regulator output should not be a problem. While protection can be added to the pre-regulator, it's no longer a simple circuit because of the additional parts needed. In service, a short is highly unlikely unless there is a catastrophic failure somewhere, and many other parts will need to be replaced anyway. Adding a couple of cheap transistors/ MOSFETs to the repair bill is the least of your problems.

Page Created and Copyright © Rod Elliott 03 Jul 2003./ Updated Nov 2012 - added Figure 2 and text to suit./ Dec 2016 - added MOSFET version./ July 19 - Clarified formulae./ Aug 23 - added Fig 3A and text.

V2 = V1 - V_zener	(Where V1 is amplifier supply voltage, and a V_zener is the zener voltage used)
R1 = R3 = V2 / ( I_zener + I_Base )	(R1 and R3 values are in kΩ, I_zener is zener current, I_Base is base current ... both in milliamps)
P = V2² / R1	(P is power dissipation of R1 and R3 in mW)

Simple Pre-Regulator